AI Paper Atlas

Research map / Video / Action recognition

Uncertainty-Boosted Robust Video Activity Anticipation

Zhaobo Qi, Shuhui Wang, Weigang Zhang, Qingming Huang

TPAMI 2024 · Video › Action recognition · approach: Action anticipation and latent actions · 13 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add uncertainty-boosted learning to RULSTMRULSTM · EK100 validation (Act)top-5 recall (%): 14.0 -> 16.6Improved
add uncertainty-boosted learning to DCRDCR · EK100 validation (Act)top-5 recall (%): 18.3 -> 19.1Improved
add uncertainty-boosted learning to ActionBanksActionBanks · EK55 validationtop-5 accuracy (%): 28.6 -> 30.2Improved
add distribution adjustment using uncertainty (ablation)Ub-Baseline · EK55 validation (1.0s)top-5 accuracy (%): 31.6 -> 32.1Improved
add precise target activity label construction (ablation)Ub-Baseline · EK55 validation (1.0s)top-5 accuracy (%): 31.2 -> 32.1Improved
add sample-wise relative uncertainty learning (ablation)Ub-Baseline · EK55 validation (1.0s)top-5 accuracy (%): 32.0 -> 32.1Improved
add temporal relative uncertainty learning (ablation)Ub-Baseline · EK55 validation (1.0s)top-5 accuracy (%): 31.6 -> 32.1Improved
add class-balanced loss to DCRDCR · EK100 validation (Act)top-5 recall (%): 14.6 -> 12.7Hurt
use uncertainty-based labels instead of class-balanced loss for tail classesUb-DCR · EK100 tail classes (Act)top-5 recall (%): 11.4 -> 13.5Improved
add Gaussian noise to features (eta 0 to 10)Ub-DCR · EK55 validationtop-5 accuracy (%): 34.2 -> 11.4Hurt
add Monte Carlo Dropout model uncertainty to Ub-DCRUb-DCR · EK55 validationtop-5 accuracy (%): 34.2 -> 34.7Improved
add Monte Carlo Dropout model uncertainty to Ub-DCRUb-DCR · EK100 validation (TSN)top-5 accuracy (%): 14.8 -> 14.7No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy